Canali Luca, Gaino Francesca, Costantino Andrea, Guizzardi Mathilda, Carnicelli Giorgia, Gullà Federica, Russo Elena, Spriano Giuseppe, Giannitto Caterina, Mercante Giuseppe
Department of Biomedical Sciences, Humanitas University, Via Rita Levi Montalcini 4, Milan, Pieve Emanuele 20072, Italy; Otorhinolaryngology Unit, IRCCS Humanitas Research Hospital, Via Manzoni 56, Milan, Rozzano 20089, Italy.
Department of Otolaryngology Head and Neck Surgery, AdventHealth Orlando, 410 Celebration Place, Celebration, Florida 34747, USA.
Auris Nasus Larynx. 2024 Dec;51(6):922-928. doi: 10.1016/j.anl.2024.09.002. Epub 2024 Sep 20.
Approximately 30 % of thyroid nodules yield an indeterminate diagnosis through conventional diagnostic strategies. The aim of this study was to develop machine learning (ML) models capable of identifying papillary thyroid carcinomas using preoperative variables.
Patients with thyroid nodules undergoing thyroid surgery were enrolled in a retrospective monocentric study. Six 2-class supervised ML models were developed to predict papillary thyroid carcinoma, by sequentially incorporating clinical-immunological, ultrasonographic, cytological, and radiomic variables.
Out of 186 patients, 92 nodules (49.5 %) were papillary thyroid carcinomas in the histological report. The Area Under the Curve (AUC) ranged from 0.41 to 0.61 using only clinical-immunological variables. All ML models exhibited an increased performance when ultrasound variables were included (AUC: 0.95-0.97). The addition of cytological (AUC: 0.86-0.97) and radiomic (AUC: 0.88-0.97) variables did not further improve ML models' performance.
ML algorithms demonstrated low accuracy when trained with clinical-immunological data. However, the inclusion of radiological data significantly improved the models' performance, while cytopathological and radiomics data did not further improve the accuracy.
Level 4.
通过传统诊断策略，约30%的甲状腺结节诊断结果不确定。本研究旨在开发能够利用术前变量识别甲状腺乳头状癌的机器学习（ML）模型。
对接受甲状腺手术的甲状腺结节患者进行一项回顾性单中心研究。通过依次纳入临床免疫、超声、细胞学和放射组学变量，开发了6种二分类监督式ML模型来预测甲状腺乳头状癌。
在186例患者中，组织学报告显示92个结节（49.5%）为甲状腺乳头状癌。仅使用临床免疫变量时，曲线下面积（AUC）范围为0.41至0.61。纳入超声变量后，所有ML模型的性能均有所提高（AUC：0.95 - 0.97）。加入细胞学（AUC：0.86 - 0.97）和放射组学（AUC：0.88 - 0.97）变量并未进一步提高ML模型的性能。
用临床免疫数据训练时，ML算法显示出较低的准确性。然而，纳入放射学数据显著提高了模型的性能，而细胞病理学和放射组学数据并未进一步提高准确性。
4级。
